Connect Unless to Claude and ChatGPT in one click

November 4, 2026

Summary You can now connect Unless to Claude, ChatGPT, and other AI apps in one click, using your normal Unless login instead of an API key. From there, you manage your Unless account by asking your AI app, with exactly the permissions you already have in the dashboard. Your Customer Agent can also do more in your own systems: give it your API documentation, and it works out which requests to make, within the limits you set.

The rest of this release helps you keep the agent’s knowledge accurate and show what it delivers. The Content Library now flags articles that no source supports anymore, and quality reports show what changed since the last run and why an answer scored low. A new monthly report sums up usage, maturity, and value for your management, and new Flex accounts get a checklist that shows what to set up first.

New Features

• Connect Unless to Claude and ChatGPT. Add Unless as a connector in Claude, ChatGPT, or another AI app that supports MCP, sign in with your normal Unless login, and manage your account by asking. The connection never gets more permissions than you have in the dashboard, and you no longer need to copy an API key.
• Connect your own API from its documentation. Paste a link to your API documentation and Unless sets up a Procedure the agent can use to look things up and make changes in your systems. You review everything before saving, you choose which actions are allowed, and the agent only calls your API for signed-in visitors.
• A monthly report for your management. On the first of every month, Unless creates a report on your AI usage, maturity, and business value, ready to share with the people who decide on your AI investment.
• A getting-started checklist for Flex. New self-service Flex accounts now see a checklist in the dashboard menu, from connecting your website to inviting a teammate, and each step ticks itself off when you complete it.
• Translate a transcript in the dashboard. Pick a language on an interaction’s details page and the full transcript is translated in place, so you can review any interaction in a language you read, at no extra cost.
• Find content that has lost its source. Every article in your Content Library now shows how reliable it is, based on whether its sources still exist. When an article loses its sources, you get a suggestion in your Inbox to remove it, and nothing is deleted without your approval.

Improvements

• Keep content out of AI answers without deleting it. Exclude an article, or a whole collection at once, from answers, search, and Help Center generation while it keeps updating from its source. You can include it again at any time.
• Compare a quality report with the previous one. Each report now shows the score change and which questions were fixed, got worse, or are new, and you can filter on those changes.
• See why an answer scored low. Low-scoring answers in a quality report now have an Analyze button that explains why the agent’s answer differs from the control answer.
• Show Help Center categories to the right audience. Limit a category to one or more audiences, so each group of users only sees the categories meant for them. Restricted categories never appear on your public Help Center.
• Get an email when a task is assigned to you. When someone assigns you a task in the dashboard Inbox, you now get an email with a direct link to it. You can turn these emails off under Settings > Notifications.
• Choose what counts in your Inbox badge. Pick which message types count toward the Inbox badge (signals, content, or AI maturity), so the number reflects the work you care about. Tasks assigned to you always count.
• Run an analysis again on a flagged response. Clear an existing analysis to run it again with different settings. Any suggestions it already produced stay as they are.
• Your own icons for Help Center categories and the contact tab. Pick an icon for each Help Center category and for the contact tab in the AI component, so both match your brand.
• A smarter contact tab. Contact tab actions now follow the same visibility rules as your other actions and hide outside your configured opening hours, and you can show up to five cards.